Improvement of speech recognition for robots using blind signal separation

AutorDaniel Bicher, Olaf Kroll-Peters, Thebin Lee, Natascha Tiotuico, Mathias Wilhelm
QuelleProceedings of the 8th conference on Signal processing, computational geometry and artificial vision. 

For modern entertainment and service robots, it is desirable to establish good verbal communication with humans. This communication may easily be influenced or even completely suppressed by noise. An often neglected source of noise is that produced by the robot motors. In this paper we will show how to extract the motor noise of a robot, separate it from the desired input, and how to improve the robustness of speech recognition using Blind Source Separation.